nlrq()
我一直在尝试用(来自quantreg
包)拟合非线性分位数回归模型。但是我发现该函数对起始值非常敏感。所以我尝试使用自启动器SSexp()
。我读到的自启动功能nls()
也适用于它。但我一直收到以下错误消息:
"Error in getInitial(formula, mf) : unused argument (mf)"
我的输入代码是:
fit12 <- nlrq(visit.rate ~ SSexp(het.total, y0, b),
data = flower_mat,tau = 0.75)
我有一种模糊的想法,getInitial
以某种方式连接到创建自启动功能。但坦率地说,我不知道问题到底出在哪里。有人可以帮我解决这个问题吗?
附加信息
# R version 3.1.2 -- "Pumpkin Helmet"
# quantreg version: 5.05
# nlrwr version : 1.1-0
# toy dataset and code
x <- c(0.3,0.6,0.9,1,1.5,2, 2.1, 2.5,3, 3,5,10,11,12,14,13,17,21,23,27,30, 50)
y <- c(0,0.1,0.3,0.4,0.6,0.2,0.27,0.2,0.25,0.4,0.15,0.05,0.25,0.2,0.3,0.35,0.1,0.15,0.1,0.14,0,0)
dat <- data.frame(x,y,stringsAsFactors=FALSE)
plot(y~x)
library(quantreg)
library(nlrwr)
fit.1 <- nlrq(y ~ SSexp(x, y0, b),
data = dat,tau = 0.75)
# above mentioned error message
with(dat, SSexp(x, 0.2, 4))
# this is working
getInitial((y ~ SSexp(x, y0, b)), dat)
# showing following error:
# Error in getInitial((y ~ SSexp(x, y0, b)), data = dat) : unused argument (dat)